Cancer Incidence among Female Flight Attendants: A

JOURNAL OF WOMEN’S HEALTH
Volume 15, Number 1, 2006
© Mary Ann Liebert, Inc.
Cancer Incidence among Female Flight Attendants:
A Meta-Analysis of Published Data
ALESSANDRA BUJA, M.D., Ph.D.,1 GIUSEPPE MASTRANGELO, M.D.,1
EGLE PERISSINOTTO,1 FRANCESCO GRIGOLETTO,1 ANNA CHIARA FRIGO,1
GIUSEPPE RAUSA, M.D.,1 VALERIA MARIN, M.D.,1 CRISTINA CANOVA,1
and FRANCESCA DOMINICI2
ABSTRACT
Background: Flight attendants are exposed to cosmic ionizing radiation and other potential
cancer risk factors, but only recently have epidemiological studies been performed to assess
the risk of cancer among these workers. The aim of the present work was to evaluate the incidence of various types of cancer among female cabin attendants by combining cancer incidence estimates reported in published studies.
Methods: All follow-up studies reporting standardized incidence ratio (SIR) for cancer
among female flight attendants were obtained from online databases and analyzed. A metaanalysis was performed by applying Bayesian hierarchical models, which take into account
studies that reported SIR 0 and natural heterogeneity of study-specific SIRs.
Results: A total of seven published studies reporting SIR for several cancer types were extracted. Meta-analysis showed a significant excess of melanoma (meta-SIR 2.15, 95% posterior interval [PI] 1.56-2.88) and breast carcinoma (meta-SIR 1.40; PI 1.19-1.65) and a slight but
not significant excess of cancer incidence across types (meta-SIR 1.11, PI 0.98-1.25).
Conclusions: Although further studies are necessary to clarify the exact role of occupational
exposure, all airlines should, as some companies do, estimate radiation dose, organize the
schedules of crew members in order to reduce further exposure in highly exposed flight attendants, inform crew members about health risks, and give special protection to pregnant
women.
to be higher in members of a flight crew than in
a control group.4 However, the disruption of circadian rhythm that occurs due to flight attendants’ work shifts might reduce melatonin production,5 leading to a decreased oncostatic effect
of this hormone.6
Several epidemiological studies on cancer in female flight attendants have been completed,7–13
with some conflicting results. The sample size in
INTRODUCTION
F
LIGHT ATTENDANTS’ OCCUPATIONAL EXPOSURE to
cosmic rays corresponds to an annual radiation dose ranging from 0.2 to 5 millisievert.1
Among cosmic radiation components, neutrons
are high linear energy transfer particles that may
induce DNA damage and no radioadaptative response.2,3 Chromosomal aberration was reported
1Department
2Department
of Environmental Medicine and Public Health, University of Padua, 35128 Padua, Italy.
of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland.
98
99
CANCER INCIDENCE IN FEMALE FLIGHT ATTENDANTS
most original studies was too small to produce
robust results, and some authors tried to overcome the problem by performing meta-analysis
of cancer in women.
A pooling study on cancer mortality, published
in 2003,14 was carried out in a large cohort of European female flight attendants. The authors reported a reduction in all-site cancer mortality
(meta-standardized mortality ratio, meta-SMR
0.78, CI 0.66, 0.95) and a slight but nonsignificant
excess in mortality from breast cancer (meta-SMR
1.11, CI 0.82, 1.48). In contrast, no meta-analysis
was undertaken on cancer incidence in a similarly
large cohort of female flight attendants. In fact,
combining only two studies, one published in
1995 (on 1577 females employed in all Finnish airlines13) and one published in 1998 (on 287 retired
females12), meta-standardized incidence ratios
(meta-SIR) of 2.31 (CI 1.24, 4.30) and 1.89 (CI 1.40,
2.56) for melanoma and breast cancer, respectively, were reported by Ballard et al.15 In 2001,
Lynge aggregated observed and expected incident cancer cases of five studies, without carrying out a formal meta-analysis,16 and found
higher than expected cancer cases for melanoma
(31 vs. 15.54), breast cancer (105 vs. 74.76), other
skin cancers (6 vs. 2.28), leukemia (4 vs. 3.72), and
all-site cancers (243 vs. 209.43).
In this paper, we perform an updated and comprehensive meta-analysis of all cancer incidence
studies (including two earlier studies and those
published in 1998–2005) carried out in European
and U.S. cohorts of female flight attendants in order to provide more stable cancer risk estimates.
MATERIALS AND METHODS
AU1
Epidemiological cohort studies published in
peer-reviewed sources until February 2004 were
sought in MEDLINE, Toxline, NIOSHTIC, and
NLM Gateway bibliographic databases, using the
folllowing terms as key words: aircrew, flight attendants, neoplasms, aviation, cosmic radiation,
and epidemiology. Six independent studies7–11,13
on cancer incidence among female cabin attendants were found. The references included in the
selected papers were inspected to identify other
studies. Cancer data found by Wartenberg and
Stapleton12 in retired U.S. cabin attendants were
drawn from Ballard’s meta-analysis.15
We extracted and entered into an electronic
database the observed and expected cases of can-
cer. We included in the meta-analysis cancer sites
for which at least two studies reported at least
one observed case, and at least two expected cases
were counted in the total population.
When studies estimated an SIR equal to 0, the
results were reported in three different circumstances: (1) the expected number of cases and the
upper limit of CI of the estimated SIR were reported, (2) the upper limit of CI of the estimated
SIR was reported, and (3) no information in addition to an SIR 0 was reported (we denoted
such cases with a No in Table 2). To include in
the meta-analysis the SIR 0 for circumstances 1
and 2, we used a two-stage Bayesian hierarchical
model. In circumstance 1, we know that the observed number of cases is equal to zero, and we
have the expected number of cases. In circumstance 2, we also know that the observed number
of cases is equal to zero, but now we impute the
expected value by taking
E sup/U
where U is the upper limit of the SIR CI, and sup
is such that the probability that zero cases would
occur under a Poisson model:
P(Y 0) exp()
would be 0.025. Taking into account circumstances 1 and 2, we assumed that the observed
number of cancer cases in the study s (Ys) has a
Poisson distribution with mean s, where
log s logEs s (first stage)
and
s N(0, 2)
where 2 denotes the between-study variability of
the log(s/Es) with respect to the overall mean (second stage). Thus, under this model specification, the meta-SIR is defined as exp(). Unfortunately, we were obligated to exclude the SIR 0 for
circumstance 3 because without any information on
the expected number of cases or on the CI, we could
not make any assessment of the statistical uncertainty of the SIR. This occurred in one study for
leukemia, colo-rectal cancer, and bone cancer, in
two studies for non-Hodgkin’s lymphoma and
bladder cancer, in three studies for other skin cancers, and in four studies for cancer of the kidneys.
100
RESULTS
T1
Table 1 shows the main characteristics of the
studies included in the meta-analysis. The followup period averaged 19.3 years. The first five studTABLE 1.
Author
Pukkala et al.13
Rafnsson et al.10
Haldorsen et al.7
Linnersjö et al.9
Reynolds et al.11
Wartenberg and Stapleton12
Lynge16
an.s.,
MAIN CHARACTERISTICS
OF
ies used similar methodology regarding the
source of cohort (employers’ list of companies or
national institutions, or union of the occupational
category or both), the inclusion/exclusion criteria, and the follow-up registry (local cancer registry), Wartenberg and Stapleton’s study12 includes a small cohort of only retired workers, and
Lynge8 reported only breast cancer incidence in
flight attendants traced through the 1970 census.
All studies took into account sex, age, and calendar period when comparing observed cases with
expected cases in the general population.
Table 2 shows the distribution of incident cancer data by author and sites. Four hundred
twenty-three cases of cancer occurred in 16,635
female flight attendants using pooled data from
all studies. The SIRs reported by Wartenberg and
Stapleton12 were highest in each cancer site. Linnersjö et al.9 did not report the SIR for cancer site
with less than two cases observed. In this event,
not specified in the original studies was reported.
Table 3 shows, separately for each cancer site,
the number of the studies, posterior mean and
95% posterior region of the heterogeneity parameter (), and the posterior mean and 95% posterior interval of the meta-SIR (exp()). We found
that the meta-SIRs are significantly higher than
1.0 for melanoma and breast cancer. In addition,
we found that study-specific SIRs for melanoma
are more heterogeneous than study-specific SIRs
for breast cancer (ˆ 0.11 and ˆ 0.07, respectively).
Note that estimates of the meta-SIR and of the
heterogeneity parameter for melanoma, equal to
2.15 and to 0.11, respectively, indicate that 95%
of the true log study-specific SIRs are within the
interval log(2.15) 1.96 0.11 0.55 and log
(2.15) 1.96 0.11 0.98 or, equivalently, that
95% of the true study-specific SIRs are within the
interval 1.73 exp(0.55) and 2.66 exp(0.98).
STUDIES INCLUDED
IN THE
META-ANALYSIS
Country
Publication
year
No. of workers
Person-years
Years of follow-up
Finland
Iceland
Norway
Sweden
USA
USA
Denmark
1995
2001
2001
2003
2002
1998
1996
1,577
1,532
3,105
2,324
6,895
287
915
22,000
27,148
60,401
39,135
n.s.a
n.s.
n.s.
1967–1992
1955–1997
1953–1996
1961–1996
1988–1995
n.s.
1970–1987
not specified in the original studies.
T2
We estimated the posterior distribution of the
meta-SIR by using Bayesian hierarchical models.
Bayesian hierarchical models provide a natural
approach for combining SIR across studies, accounting for study-specific statistical uncertainty
in the estimated SIR (within study variance) and
for natural variability (heterogeneity) of the SIR
across studies (between-study variance). In addition, Bayesian hierarchical models allow us to include the studies that reported SIR 0 for circumstances 1 and 2. We fitted the Bayesian
hierarchical models by use of Monte Carlo
Markov Chain Methods, implemented by the
software WinBUGS.17,18
Because of the limited power to detect heterogeneity, we estimated the meta-SIR for the cancer types with fewer than four studies by use of
a fixed effect model that assumes that 0.19
We also performed an influential/sensitivity
analysis20 by calculating a pooled SIR only for a
subgroup of studies reporting similar methodology (such as source of cohort, inclusion/exclusion criteria, and follow-up registry used) after
excluding cancer data from Wartenberg and Stapleton,15 who studied a cohort of retired workers, and Lynge,8 who reported only breast cancer
incidence in flight attendants traced through the
1970 census.
Finally, using Stata programs,21 we investigated whether there was evidence of publication
bias by applying three methods: Begg’s test,22 Egger’s test,23 and a meta-regression model with
study size as explanatory variable.24
AU2
BUJA ET AL.
T3
1.0 0.0
5.8
1
1.1
0.0
0.3
0.3
0.3
bn.r.,
n
SIR CI
0.7 0.2
1.3 0.3
3.4
2.1
3.3
1 3.1 0.9 8.0
3
4
n
4
OF
0.9
0.0
0.3
0.0
0.0
0.8
CI
0
0
1.1
0
2.0 1.0
5.2 1.7
121 17.0
2.1 0.3
1.4
0
2.0
0
0
3.4
SIR
SIR CI
n
13.6
17.1
0
2
n.o.
0
0
6.6
0
n.r.
n.r.
2.3 17
n.r.b
61.5 0
n.r.
14.4 1
n.r.
6.0 0
n.r.
92.2 0 0
0 92.2
0
13.4 2
n.r.
n.o.
n.r.
4.3 7 1.6 0.9 2.7 14
16.6 3
n.r.
859.0 1
n.r.
14.6 1
n.r.
n.o.
n.r.
n.o.
n.r.
15.4 0
n.r.
n
Lynge (1996)16
DISEASES, IX REVISION (ICD-IX)
Wartenberg and
Stapleton (1998)12
n.r. 4.3
n.r.
76
n.r.
3
3
n.r.
11
0
33
4
2
2
0
2
4
Linnersjö
et al. (2003)9
0.9 1.3 127 1.0 0.8 1.2
0.0 1.8
0
0.7 2.1 12 0.6 0.1 1.9
0.5 2.9
6 1.1 0.2 3.2
0.0 13.0
0
1.0 2.7 19 2.2 1.1 3.9
1.0 6.9
5 0
0.0 4.35
0.8 1.5 38 1.3 0.9 1.7
0.7 2.0 16 0.7 0.2 1.8
0.1 1.6
3 0.6 0.1 2.3
0.3 1.7
6 0.4 0.1 1.5
0.2 5.2
2 0
0.0 3.9
0.0 3.4
1 1.7 0.2 6.0
0.0 1.1
1 1.1 0.3 2.7
n.o. 0.6 0.0
1
2
1.1
0
1.3
1.3
0
1.7
2.9
1.1
1.2
0.5
0.8
1.4
0.6
0.2
SIR CI
Haldorsen
et al. (2001)7
TUMOR SITE: INTERNATIONAL CLASSIFICATION
standardized incidence ratio; CI, 95% confidence interval; n, number of observed cases.
none reported.
cn.o., none observed.
dBasal cell is not included.
aSIR,
2
6
2
3.6 0.4 12.9
3.4
7.5
204–208
1 1.6 0.6
n.o. 2.1 0.2
0.3 0.0 1.4
0.8 0.1 2.8
3.4
0.6 0.0
1.4
0.3
1.0
0.9
3.3
n
AND
1.3 104
5.8
1
2.2
5
1.2
3
n.o.
4.4 12
n.o.
1.8 60
1.2
2
2.4
5
2.1
5
12.1
2
n.o.
3.6
1
193
200,202
0.9
0.0
0.3
0.1
2.5 1.3
1.1
1.0
0.9
0.4
0.6 0.0
n
140–208 1.2 0.9 1.7 35 1.2 1.0 1.6 64
151
1.0 0.0 5.8 1 1.2 0.0 6.6
1
153,154 1.3 0.2 4.8 2
n.o.c
162
1.6 0.0 9.0 1 0
0.0 1.1
0
170
15.1 1.8 54.4 2 4.3 0.1 23.3
1
172
2.1 0.4 6.2 3 3.0 1.2 6.7
7
173
n.o. 1.7 0.0 9.7
1
174
1.9 1.2 2.2 20 1.5 1.0 2.1 26
180
0
0.0 3.7 0 0.9 0.2 2.3
4
182
0
0.0 3.0 0 1.5 0.3 4.3
3
183
0.5 0.0 2.6 1 0.8 0.2 2.4
3
188
n.o. 2.0 0.0 10.8
1
189
n.o.
n.o.
191,192 0.5 0.0 2.9 1 1.4 0.3 4.0
3
SIR CI
PUBLICATION)
Reynolds
et al. (2002)11
OF
SIR CI
AUTHOR (YEAR
Rafnsson
et al. (2001)10
BY
All sites
Stomach
Colon/rectum
Lung
Bone
Melanoma
Other skind
Breast
Cervix uteri
Body of uterus
Ovary
Bladder
Kidney
Brain/nervous
system
Thyroid
Non-Hodgkin’s
lymphoma
Leukemia
n
CANCER DATA
Pukkala
et al. (1995)13
OF INCIDENT
ICD-IX SIRa CI
DISTRIBUTION
Cancer sites
TABLE 2.
102
BUJA ET AL.
TABLE 3. AGGREGATED CANCER DATA BY SITE OF TUMOR:
INTERNATIONAL CLASSIFICATION OF DISEASES, IX REVISION (ICD-IX)
Cancer site
ICD-IX
na
PI of Meta-SIR
PI of meta-SIR
All sites
Stomach
Colon/rectum
Lung
Melanoma
Other skinb
Breast
Cervix uteri
Body of uterus
Ovary
Bladder
Kidney
Brain/nervous system
Thyroid
Non-Hodgkin’s lymphoma
Leukemia
140–208
151
153, 154
162
172
173
174
180
182
183
188
189
191, 192
193
200, 202
204–208
6
5
5
6
6
3
7
6
6
6
4
2
6
5
3
6
0.05
0.39
0.14
0.53
0.11
n.c.c
0.07
0.49
0.25
0.14
0.39
n.c.
0.29
0.37
n.c.
0.43
0.01–0.19
0.01–2.28
0.01–0.64
0.01–2.61
0.01–0.44
1.11
0.60
1.06
0.66
2.15
1.91
1.40
0.89
0.84
0.74
1.45
1.05
0.65
0.92
1.19
1.43
0.99–1.25
0.09–1.53
0.64–1.60
0.15–1.27
1.56–2.88
0.71–3.73
1.19–1.65
0.34–1.56
0.39–1.46
0.41–1.15
0.33–3.16
0.22–2.53
0.24–1.20
0.35–1.67
0.52–2.15
0.33–2.86
0.01–0.27
0.01–1.84
0.01–1.58
0.01–0.63
0.01–2.26
0.01–1.41
0.01–1.74
0.01–2.15
an, number of studies; , posterior mean of the heterogeneity parameter; PI, 95% posterior interval; meta-SIR, posterior mean of the meta-SIR.
bBasal cell is not included.
cn.c., heterogeneity not calculated because of the small number of studies.
After excluding studies by Wartenberg and Stapleton12 and by Lynge,8 meta-SIRs reported in
Table 3 were only marginally affected (data not
shown).
We investigated publication bias by using SIRs
for breast cancer, the only cancer site available for
all the seven studies included in the meta-analysis. No publication bias was detected by any of
the tests used (Begg’s test, p 0.37; Egger’s test,
p 0.48; meta-regression coefficient for the explanatory variable (study size), p 0.69).
DISCUSSION
In a commentary (that includes all cohorts except those from Linnersjö et al.9 and Reynolds et
al.,11 Lynge16 found higher than expected cancer
cases for leukemia, melanoma, other skin tumors,
and breast and all-site tumors. Likewise, in an
earlier analysis using a standard meta-analytical
approach with random effects,25 we evidenced a
significant excess for leukemia, melanoma, other
skin tumors, and breast and all-site cancer incidence (data not shown). This approach led to an
overestimation of the pooled SIRs (because it excludes studies reporting SIR 0 ) and too narrow
CI (because it assumes that the between-study
variance is known, when in fact it is estimated
from the data).
The Bayesian hierarchical approach does not
exclude studies reporting SIR 0 and gives
wider CIs because it automatically takes into account the extra uncertainty due to the betweenstudy variance. Analysis using Bayesian methods
allowed estimation of a significant incidence excess for melanoma (meta-SIR 2.15, 95% posterior interval [PI] 1.56-2.88) and breast carcinoma
(meta-SIR 1.40; PI 1.19-1.65). A slight but not
significant excess (meta-SIR 1.11, PI 0.98-1.25)
was found for all site tumors. Our findings agree
with those reported by Ballard et al.15
In the multicenter study on mortality,14 the risk
of death from breast cancer was 1.11 (CI 0.820.48) and for melanoma was 0.36 (CI 0.04-1.37),
whereas in the present study, the meta-SIR denotes a 40% excess (PI 1.19-1.65) and 115%
(PI 1.56-2.88) excess, respectively, for breast
cancer and melanoma in female flight attendants,
with respect to the general population (Table 3).
These differences can be explained by intensive
medical surveillance, which increases incidence
but reduces mortality. For tumors with a high
chance of recovery after early diagnosis, such as
melanoma or breast cancer, however cancer incidence is an indicator of increased risk more than
of mortality. We, therefore, undertook the present
work, which is, to date, the first comprehensive
meta-analysis on cancer incidence among U.S.
and European female flight attendants.
serted that after taking into account the reproductive history, the risk of breast cancer is unlikely to
be explained solely by confounding due to parity.
Pukkala et al.13 found that after the differences in
reproductive and other factors related to social
class were taken into account, a pronounced excess significant risk of breast cancer remains only
15 years after recruitment. Accordingly, in a recent
case-control study, confounding factors (social
class, reproductive factors) did not explain the
magnitude of excess risk for breast cancer among
cabin attendants.30 Furthermore, an independent
effect between radiation exposure and reproductive history on breast cancer risk was reported
among survivors of the atomic bomb.31 However,
an occupational origin for breast cancer is supported by the linear dose-response relationship
found between radia-tion exposure and breast cancer risk in a pooled analysis of eight cohorts of differently exposed women.32
Melanoma and breast cancer excesses might be
attributed to radiation exposure largely on the basis of analogy with low-LET radiation-exposed
populations. Much less is known about biological effects of low-dose exposure to high-LET than
to low-LET. At high altitudes, however, about
half of the effective dose would be due to highLET neutrons rather than low-LET gamma rays.33
High-LET exposure confers more biological damage than low-LET, mainly because of the higher
amount of ionizations that occur in the tissue. The
relative biological effectiveness (RBE), the ratio of
dose of low-LET gamma-radiation/dose of the
high-LET radiation of interest, that caused the
same biological effect seems to range from 5 to 20
for neutrons depending on the dose and energy
of the neutron exposure. Furthermore, for neutrons the RBE increased with decreasing dose, so
RBE would be expected to be close to 20 for the
low-dose exposure of the air crew.34
A desynchronized production of melatonin by
the pineal gland was found in flight attendants,
because of disruption of the circadian rhythm.5
Melatonin was found to have oncosuppressive
properties on melanoma cells35,36 and an oncostatic effect on mammary glands both in vivo and
in vitro.37,38 Melatonin could also act as a naturally
occurring antiestrogen, thereby influencing the
proliferative rate of mammary tumor cells.39,40
Epidemiological evidence confirmed a higher rate
of breast cancer with increasing duration of nighttime employment41,42 and degree of visual impairment in women.43–45
AU3
Occupational cohorts benefit from improved
medical surveillance and greater access to diagnostic technologies with respect to the general
population. This diagnostic sensitivity bias has
increased over time as employers, employees,
and physicians have become more aware of
work-related diseases. This bias might increase
the incidence of cancer. On the other hand, any
epidemiological study on cohorts of workers is
unavoidably affected by the selection bias known
as healthy worker effect (HWE), which presumably derives from a screening process, perhaps a
largely self-selection one, that allows relatively
healthy people to become or remain workers. The
bias arises when workers are compared with the
general population, where those who remain unemployed, retired, disabled, or otherwise out of
active service are a less healthy group. Bias from
HWE is always in the direction of null hypothesis. Therefore, although the net balance is unpredictable, the opposing direction forms of the biases may have cancelled each other out.
A bias may also have occurred because of missing data. Linnersjö et al.9 did not report the SIR
for cancer site with less than two cases observed
and, therefore, could not be included in the metaanalysis. This could be a publication bias. However, the presence of a publication bias was excluded by the Begg’s test, the more sensitive
Egger’s test, and the meta-regression approach,
which was applied because the sensitivity of the
former two methods is generally low in metaanalyses based on less than 20 trials.23
The association between exposure and melanoma risk might be confounded by intermittent
sun exposure, as the frequency of traveling to
sunny countries may be greater in flight attendants than in the general population. In a recent
case-control study, however, Rafnsson et al.26
concluded that the increased incidence of malignant melanoma in female cabin attendants cannot be solely explained by excessive sun exposure. Supporting evidence of an association
between melanoma and radiation comes from
studies carried out in nuclear industry workers,27
patients undergoing radiotherapy,28 and U.S. radiology technologists exposed to particularly
high radiation doses.29
Linnersjö et al.9 estimated that in female cabin
crew, reproductive history could yield a 10% increase in breast cancer incidence, which estimation
does not seem to fully explain the cancer excess
observed in the same workers. Rafnsson et al.9 as-
103
CANCER INCIDENCE IN FEMALE FLIGHT ATTENDANTS
AU4
104
BUJA ET AL.
CONCLUSIONS
Despite the relatively short follow-up period
(19.3 years per study), meta-SIRs for melanoma
and breast cancer were significantly increased in
flight attendants with respect to the general population. Although further studies are necessary to
clarify the exact role of occupational exposure, all
airlines should, as some companies do, estimate
radiation dose, organize the schedules of crew
members in order to reduce further exposure in
highly exposed flight attendants, inform them
about health risks, and give special protection to
pregnant women.
ACKNOWLEDGMENTS
We are grateful to Dr. Estella Musacchio and
Ms. Linda Inverso for their help in revising the
manuscript.
REFERENCES
AU5
1. Waters M, Bloom TF, Grajewski B. The NIOSH/FAA
Working Women’s Health Study: Evaluation of the
cosmic-radiation exposures of flight attendants. Federal Aviation Administration. Health Phys 2000;79:
553.
2. Friedberg W, Copeland K, Duke FE, et al. Radiation
exposure of aircrew. In: Green KB, ed. The aviation
industry. Philadelphia: Hanley & Belfus, Inc., 2002:
293.
3. Onhishi T, Takahashi A, Ohnishi K. Studies about
space radiation promote new fields in radiation biology [Review]. J Radiat Res (Tokyo) 2002;43(Suppl):7S.
4. Romano E, Ferrucci L, Nicolai F, et al. Increase of
chromosomal aberrations induced by ionising radiation in peripheral blood lymphocytes of civil aviation
pilots and crew members. Mutat Res 1997;377:89.
5. Grajewski B, Nguyen MM, Whelan EA, Cole RJ, Hein
MJ. Measuring and identifying large-study metrics
disruption circadian rhythm in female flight attendants. Scand J Work Environ Health 2003;29:337.
6. Stevens, Richard G. Circadian disruption and breast
cancer: From melatonin to clock genes. Epidemiology
2005;16:254.
7. Haldorsen T, Reitan JB, Tveten U. Cancer incidence
among Norwegian airline cabin attendants. Int J Epidemiol 2001;30:825.
8. Lynge E. Risk of breast cancer is also increased among
Danish female airline cabin attendants. BMJ 1996;
312:253.
9. Linnersjö A, Hammar N, Dammstrom BG, et al. Cancer incidence in airline cabin crew: Experience from
Sweden. Occup Environ Med 2003;60:810.
10. Rafnsson V, Tulinius H, Jonasson JG, Hrafnkelsson J.
Risk of breast cancer in female flight attendants: A
population-based study (Iceland). Cancer Causes
Control 2001;12:95.
11. Reynolds P, Cone J, Layefsky M, et al. Cancer incidence in California flight attendants (United States).
Cancer Causes Control 2002;13:317.
12. Wartenberg D, Stapleton CP. Risk of breast cancer is
also increased among retired U.S. female airline cabin
attendants. BMJ 1998;316:1902.
13. Pukkala E, Auvinen A, Wahlberg G. Incidence of cancer among Finnish airline cabin attendants, 1967–92.
BMJ 1995;311:649.
14. Zeeb H, Blettner M, Langner I, et al. Mortality from
cancer and other causes among airline cabin attendants in Europe: A collaborative cohort study in eight
countries. Am J Epidemiol 2003;158:35.
15. Ballard T, Lagorio S, De Angelis G, Verdecchia A.
Cancer incidence and mortality among flight personnel: A meta-analysis. Aviat Space Environ Med 2000;
71:2176.
16. Lynge E. Commentary: Cancer in the air. Int J Epidemiol 2001;30:830.
17. Spiegelhalter DJ, Thomas A, Best NG, et al. WinBUGS
Version 1.4 User Manual. MCR Biostatistic Unit, 2003.
18. Smith TC, Spiegelhalter DJ, Thomas A. Bayesian approaches to random-effects meta-analysis: A comparative study. Stat Med 1995;14:2685.
19. Leonard T, Duffy JC. A Bayesian fixed effects analysis of the Mantel-Haenszel model applied to metaanalysis. Stat Med. 2002;21:2295.
20. Egger M, Smith GD, Phillips AN. Meta-analysis: Principles and procedures. BMJ 1997;315:1533.
21. Sharps SJ, Sterne JAC. Sbe16: Meta-analysis. Stata
Tech. Bull 1997;38:9.
22. Begg CB, Mazumdar M. Operating characteristics of
a rank correlation test for publication bias. Biometrics
1994;50:1088.
23. Egger M, Smith GG, Schneider M, Minder C. Bias in
meta-analysis detected by a simple, graphical test.
BMJ 1997;315:629.
24. Sterne JAC, Egger M, Smith D. Systematic reviews in
health care: Investigating and dealing with publication and other biases in meta-analysis. BMJ 2001;
323:101.
25. DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clin Trials 1986;7:177.
26. Rafnsson V, Hrafnkelsson J, Tulinius H, et al. Risk factors for cutaneous malignant melanoma among aircrews and a random sample of the population. Occup
Environ Med 2003;60:815.
27. Austin DF, Reynolds P. Investigation of an excess of
melanoma among employees of the Lawrence Livermore National Laboratory. Am J Epidemiol 1997;145:
524.
28. Waxler B. Malignant melanoma in patient who received neutron beam therapy. Lancet 1984;21:904.
29. Freedman DM, Sigurdson A, Rao RS, et al. Risk of
melanoma among radiologic technologists in the
United States. Int J Cancer 2003;103:556.
CANCER INCIDENCE IN FEMALE FLIGHT ATTENDANTS
30. Rafnsson V, Sulem P, Tulinius H, Hrafnkelsson J.
Breast cancer risk in airline cabin attendants: A nested
case-control study in Iceland. Occup Environ Med
2003;60:807.
31. Land CE, Hayakawa N, Machado SG, et al. A casecontrol interview study of breast cancer among Japanese A-bomb survivors. II. Interactions with radiation
dose. Cancer Causes Control 1994;5:167.
32. Preston DL, Mattsson A, Holmberg E, et al. Radiation
effects on breast cancer risk: A pooled analysis of
eight cohorts. Radiat Res 2002;158:220.
33. Boice JD Jr, Blettner M, Auvinen A. Epidemiologic
studies of pilots and aircrew. Health Phys 2000;79:576.
34. Sigurdson AJ, Ron E. Cosmic radiation exposure and
cancer risk among flight crew. Cancer Invest 2004;22:743.
35. Ying SW, Niles LP, Crocker C. Human malignant
melanoma cells express high-affinity receptors for
melatonin: Antiproliferative effects of melatonin and
6-chloromelatonin. Eur J Pharmacol 1993;246:89.
36. Souza AV, Visconti MA, De Lauro Castrucci AM.
Melatonin biological activity and binding sites in human melanoma cells. J Pineal Res 2003;34:242.
37. Mediavilla MD, Guemez A, Ramos S, et al. Effects of
melatonin on mammary gland lesions in transgenic
mice overexpressing N-ras proto-oncogene. J Pineal
Res 1997;22:86.
38. Cos S, Mediavilla MD, Fernandez R, et al. Does melatonin induce apoptosis in MCF-7 human breast cancer cells in vitro? J Pineal Res 2002;32:90.
39. Souza AV, Visconti MA, De Lauro Castrucci AM.
Melatonin biological activity and binding sites in human melanoma cells. J Pineal Res 2003;34:242.
105
40. Anisimov VN. The role of pineal gland in breast cancer development. Crit Rev Oncol Hematol 2003;
46:221.
41. Hansen J. Increased breast cancer risk among women
who work predominantly at night. Epidemiology
2001;12:74.
42. Davis S, Mirick DK, Stevens RG. Night shift work,
light at night, and risk of breast cancer. J Natl Cancer
Inst 2001;93:1557.
43. Verkasalo PK, Pukkala E, Stevens RG, et al. Inverse
association between breast cancer incidence and degree of visual impairment in Finland. Br J Cancer
1999;80:1459.
44. Kliukiene J, Tynes T, Andersen A. Risk of breast cancer among Norwegian women with visual impairment. Br J Cancer 2001;84:397.
45. Feychting M, Osterlund B, Ahlbom A. Reduced cancer incidence among the blind. Epidemiology 1998;
9:490.
Address reprint requests:
Alessandra Buja, M.D., Ph.D.
Department of Environmental Medicine and
Public Health
University of Padua Via Loredan 18
35128 Padua
Italy
E-mail: [email protected]
KBUJA ET AL.
BUJA
AU1
Meaning of sentence starting with “Cancer” not clear. Please
clarify
AU2
Ref 15 is Ballard et al. Please clarify & correct
AU3
Define LET
AU4
Clarify meaning of sentence starting with “Furthermore”.
AU5
Title correct? ref. 5